Momentum AI

Momentum democratizes AI by providing no-coding toolkits to rapidly train and deploy ML models in production, thus increasing the overall productivity of the data science team.

No specialized skill is needed to work with Momentum.

request demo
Momentum Works

How Does Momentum AI Work?

  • Prepare your data using Momentum Connect.
  • Select a model type and configure it to take your data as input.
  • Execute to automatically train a number of algorithms suitable for your use case.
  • Evaluate and select the best model.
  • Push the model to deploy it in production.
  • Monitor model performance over time to assess when the right time to retrain the model is.
  • Retrain incrementally, if necessary, and maintain model versions to switch to the best performing models.

Supported Algorithms

Momentum AI supports a wide variety of machine learning algorithms out of the box. It provides a pluggable architecture to add new algorithms with just a few simple configurations. Here is a high-level list of models that we support:

Supervised Regression

  • Generalized Linear Regression
  • Linear Regression
  • Random Forest Regression
  • Decision Tree Regression
  • Deep Learning/ANN Regression
  • String to Index Model
  • Recurrent Neural Network Regression(LSTM)
  • Gradient-Boosted Tree (GBT) Regression
  • Survival Regression
  • Isotonic Regression
  • Factorization Machines Regression

Supervised Classification

  • Logistic Regression
  • Decision Tree Classifier
  • Random Forest Classifier
  • Deep Learning/ Artificial Neural Network/
    Multilayer Perceptron Classifier
  • Markov Chain with Neural Network
  • Convolutional Neural Network (CNN)
  • Gradient-Boosted Tree (GBT) Classifier
  • Linear Support Vector Machine (LSVM)
  • Naive Bayes Classifier
  • Factorization Machines Classifier

Unsupervised Machine Learning

  • K-Means Clustering
  • Latent Dirichlet Allocation (LDA) Clustering
  • Bisecting K-means Clustering
  • Gaussian Mixture Model (GMM) Clustering
  • Power Iteration Clustering (PIC)

Natural Language Processing (NLP)

  • Word2Vec
  • Document Similarity
  • Tokenization, Sentence segmentation, POS, NER
    and concept categorization
  • Text Summarization
  • Sentiment Analysis

Recommender Engine / Collaborative Filtering using Alternating Least Squares

Computer Vision

  • LSTM for OCR and ICR
  • Convolutional Neural Network (CNN)
  • Object Detection Using Single Shot Multibox
    Detection (SSD)
  • Object Detection Using YOLO
  • Object Detection Using RCNN, Fast RCNN, and
    Faster RCNN
  • Facial Recognition

Feature Engineering

  • Pearson’s Chi-squared
  • Correlation Coefficient – Pearson and Spearman
  • SMOTE
  • String to Index
  • OneHotEncoder
  • Imputer
  • PCA

Feature Engineering

Momentum AI does automatic feature engineering. This helps data scientists to stay focused on improving the model accuracy.

You can also perform feature engineering using box plotting, Pearson’s Chi-squared, and correlation coefficients.

If your dataset contains unbalanced classes and downsampling will cause data losses, you could use SMOTE to create synthetic data to balance your classes. Momentum provides a highly scalable SMOTE implementation to work with billions of data rows.

Model Deployment

Efficient versioning, management, and deployment of machine learning models are essential for any successful AI implementation. Momentum provides features to easily manage and deploy AI models in production in one of the following deployment use cases:

  • Deploy model for realtime prediction from streaming data
  • Deploy model as a Restful webservice
  • Deploy model for on-demand prediction from a batch of data
  • Deploy one or more models to work together to automate business processes. Learn more about Momentum Automate
See MLOps for more details

Ready To Embrace The Future

If you are working on a data engineering or AI solution, trying to explore a use case, or building a proof-of-concept, please contact us for a one-on-one discussion.

request demo